Redes neuronales para predecir el comportamiento del conjunto de activos financieros más líquidos del mercado de valores peruano

• Autores: Bellido Anicama, Alfredo Bruno, Schwarz Díaz, Max
• Formato: artículo
• Fecha de Publicación: 2019
• Institución: Universidad de Lima
• Repositorio: ULIMA-Institucional
• Lenguaje: español
• OAI Identifier: oai:repositorio.ulima.edu.pe:20.500.12724/8256
• Enlace del recurso: https://hdl.handle.net/20.500.12724/8256; https://dx.doi.org/10.18004/ucsa/2409-8752/2019.006(01)049-064.
• Nivel de acceso: acceso abierto
• Materia: Financial risk; Capital market; Artificial intelligence; Neural networks (Computer science); Riesgo financiero; Mercado de capitales; Inteligencia artificial; Redes neuronales (Informática); https://purl.org/pe-repo/ocde/ford#5.02.04

The purpose of this research is to identify an artificial intelligence tool based on neural networks to predict the behavior of performance and risk of the set of financial assets based on actions that more accurately reflect the stock market movement of the Peruvian stock market. The research initially identified the most appropriate financial asset to estimate the performance and risk values of the 50% most liquid share portfolio in the Peruvian market in the 2010-2016 period. From the selected asset, the technique of artificial neural networks with a multilayer perceptron with regression configured with 3 layers (21,85,2) was used, using a logistic activation function with an LBFGS optimizer at a learning rate of 0.01 to establish the financial, operational, commercial or corporate governance patterns that can explain and / or predict the behavior of the same in the market. The research concludes that the cash generation capacity and the speed with which the assets are rotated, as well as the speed with which the Capex is disbursed, constitute the main factors that influence the determination of the best combinations of performance and risk for the group of financial assets considered as a subject of study, independent of the market sector in which it operates. The research found a neural network able to approximate the prediction of performance and risk with a 76.93% efficiency for the set of assets selected in the study period. The research provides a recognition of differentiated patterns in financial, operational, commercial and corporate governance aspects with a special emphasis on the managerial capacity that generates them whose influence is reflected in the performance of the set of assets studied through the technique of neural networks generating a predictive tool to estimate its stock market behavior.